Quite. If I was in a position to be firing people, though, I'd be shooting for considerably more than 95% confidence -- a level of confidence that should be achievable with the kind of data sets that the OP was talking about.
What proportion of false positives would you shoot for? The current status quo seems to be towards 0% false positive, and it appears that some areas might actually reach that goal.
Post will be returning in Main, after a rewrite by the company's writing staff. Citations Galore.